Abstract

   Neuromyelitis optica spectrum disorder (NMOSD) is an autoimmune disease
   characterized by anti‐aquaporin 4 (AQP4) antibody‐mediated astrocyte
   damage and subsequent demyelination. Prior attempts to treat NMOSD with
   interferon‐beta (IFN‐β), a disease‐modifying therapy for multiple
   sclerosis, resulted in worsening of disease activity, with an unknown
   mechanism. Here, robust activation of the cGAS‐STING‐IFN‐I signaling
   pathway is identified in myeloid cells in both the periphery and
   central nervous system. The abnormal IFN‐I response gives rise to an
   increase in the number of AQP4 antigen‐specific autoreactive T cells.
   Sting deficiency can significantly blunt the activation of
   AQP4‐specific T cells, as well as the IFN‐I activity in microglia, and
   attenuate astrocyte damage. Consequently, the clinical manifestation of
   NMOSD is ameliorated in a passive transfer mouse model of NMOSD.
   Further, treatment with STING inhibitor H151 alleviates the severity of
   NMOSD mouse models. These findings uncover the cGAS‐STING‐IFN‐I pathway
   in promoting autoreactive T cells and establish a foundation for
   inhibiting this pathway as a new therapeutic revenue for NMOSD.

   Keywords: innate immune response, neuromyelitis optica spectrum
   disorder, T cells, type I interferon
     __________________________________________________________________

   The cGAS‐STING innate immune pathway is excessively activated both in
   peripheral monocytes and CNS microglia in patients with NMOSD, leading
   to an overproduction of IFN‐I. IFN‐I subsequently acts on AQP4‐specific
   autoreactive T cells. Blocking this pathway may offer novel therapeutic
   options for individuals with NMOSD.

   graphic file with name ADVS-12-e00942-g005.jpg

1. Introduction

   Neuromyelitis optica spectrum disorder (NMOSD) is an autoimmune disease
   affecting the central nervous system (CNS), characterized by the
   presence of autoantibodies against aquaporin‐4 (AQP4‐IgG) in majority
   of patients.^[ [50]^1 ^] AQP4‐IgG, generated by autoreactive B cells,
   directly damages astrocytes through binding to astrocytic endfeet.^[
   [51]^2 ^] Several studies have highlighted the crucial role of the
   innate immune response in NMOSD pathogenesis. Specifically, as CNS
   resident macrophages, microglia likely serve as primary responders to
   astrocytes upon activation by AQP4‐IgG. The presence of activated
   microglia around NMOSD lesion sites can potentially initiate innate
   immune responses, resulting in infiltration of inflammatory cells,
   complement deposition, and neuronal and oligodendrocyte damage within
   the CNS.^[ [52]^3 , [53]^4 ^]

   Interferon‐beta (IFN‐β) subtypes have been used as immunomodulatory
   agents for relapsing‐remitting multiple sclerosis (MS).^[ [54]^5 ^]
   Paradoxically, IFN‐β therapy exacerbates NMOSD activity,^[ [55]^6 ,
   [56]^7 ^]though the underlying mechanism remains elusive. Dysregulated
   type I interferon (IFN‐I) activity in the blood of NMOSD patients has
   been associated with an increase in interleukin‐6 (IL‐6) levels, which
   promote the differentiation of pathogenic Th17 cells.^[ [57]^8 ^]
   Further, aberrations in IFN‐I responses have been observed in both the
   cerebrospinal fluid (CSF) and peripheral blood B cells of NMOSD
   patients.^[ [58]^9 ^] The presence of IFN‐I also facilitates the
   differentiation of B cells into antibody‐secreting cells, thereby
   augmenting AQP4‐IgG production.^[ [59]^9 ^] However, the molecular
   mechanisms underlying abnormal IFN‐I production in NMOSD patients and
   their role in exacerbating the disease are not fully understood. This
   study aims to examine the contributions of IFN‐I to NMOSD as well as
   its molecular underpinnings, and clinical implications.

2. Results

2.1. Microglia and CNS‐Associated Macrophages Exhibited Enhanced
cGAS‐STING‐IFN‐I Activity

   Paired cerebrospinal fluid (CSF) and peripheral blood mononuclear cells
   (PBMCs) samples were collected from five NMOSD patients with
   treatment‐naive AQP4‐IgG positive NMOSD for single‐cell RNA sequencing
   (scRNA‐seq) analysis. In addition, we included a total of 8 paired CSF
   and PMBCs samples from 3 healthy controls (HC) and 5 pairs of
   non‐neuroinflammatory samples from two independent scRNA‐seq studies as
   the control group for subsequent analysis^[ [60]^10 , [61]^11 ^]
   (Figure [62]1A; and Table [63]S1, Supporting Information). After
   implementing a rigorous quality control process and addressing batch
   effects, we were able to effectively isolate 134, 603 immune cells. We
   identified main cell types including CD4^+ and CD8^+ T cells, T
   regulatory cells (Tregs), natural killer (NK) cells, B cells, plasma
   cells, conventional dendritic cells (cDCs), plasmacytoid dendritic
   cells (pDCs), as well as monocytes through our analysis (Figure
   [64]S1A–D, Supporting Information). Differential genes expression
   (DEGs) analysis over different cell types revealed that monocytes and T
   cells exhibited the most significant gene alterations in both blood and
   CSF (Figure [65]S1E,G, Supporting Information), with an overlap in
   inflammation‐related genes (Figure [66]S1F,H, Supporting Information).

Figure 1.

   Figure 1
   [67]Open in a new tab

   Enhanced cGAS‐STING‐IFN‐I signaling in CSF microglia‐like cells of
   NMOSD patients. A) Workflow illustrating the study design and sample
   selection for scRNA‐seq analysis of peripheral blood and CSF. B)
   Uniform manifold approximation and projection (UMAP) plot illustrating
   CSF monocytes subclusters based on scRNA‐seq data. C) The expression
   levels of representative marker genes used to classify microglia‐like
   cells, CNS‐associated macrophages‐like cells, and monocytes states. D)
   Correlations between CSF monocytes subclusters in CSF and three cell
   states: microglia, CNS‐associated macrophages, and monocytes. E) UpSet
   plot illustrating size and overlap of DEGs in CSF monocytes subclusters
   between NMOSD patients and controls. F) The top DEGs of the CSF
   monocyte and microglia subclusters. G) GSEA plot illustrating selected
   significantly enriched KEGG pathways in the CSF microglia‐like cells
   and CNS‐associated macrophages‐like cells between control individuals
   and NMOSD patients. H) cGAS‐STING‐IFN‐I gene‐set score in the CSF
   monocyte and microglia subclusters, comparing NMOSD patients with
   controls using the Wilcoxon rank‐sum test. I) The levels of cfDNA in
   CSF were detected from both control and patients with NMOSD. (control =
   12, NMOSD = 20). Data were analyzed using Mann Whitney test; Mean ±
   s.e.m; * p < 0.05. CAMs, CNS‐associated macrophages.

   Several studies have identified cells resembling microglia and
   CNS‐associated macrophages‐like cells in the CSF.^[ [68]^12 , [69]^13 ,
   [70]^14 ^] To further investigate the cellular alterations associated
   with neuroinflammation in CSF myeloid cells of NMOSD patients, we
   conducted re‐clustering analysis of monocytes in the CSF
   (Figure [71]1B). This analysis revealed three distinct cell subtypes
   based on their unique molecular markers: two clusters of CNS‐associated
   macrophages (C1_HAMP and C5_LYVE1, marked by LYVE1, CD163, and MRC1),
   two clusters of monocytes (C2_S100A8 and C4_INSIG1, identified by
   S100A8, CCR2 and VCAN), and two clusters of microglia (C3_CCL4 and
   C6_SPP1, characterized by P2RY12, CSF1R, and TREM2), (Figure [72]1C,D;
   Figure [73]S2A, Supporting Information). KEGG (Kyoto Encyclopedia of
   Genes and Genomes) enrichment analysis assessed the functional states
   of the identified clusters of microglia‐like and CNS‐associated
   macrophages‐like cells (Figure [74]S2B,C, Supporting Information).
   C1_HAMP displayed characteristics associated with autoimmune diseases
   and antiviral immunity terms, such as “Epstein‐Barr virus infection”
   and “Cytokine‐cytokine receptor interaction”. Other sub‐clusters, like
   C3_CCL4 and C4_INSIG1, associated with inflammatory responses,
   displayed by profiles characterized by the expression profile of
   chemokine and cytokine genes, such as CCL4, and IL1B (Figure [75]S2B,
   Supporting Information), and terms like “Th17 cell differentiation” and
   “antigen processing and presentation” (APC) in the KEGG analysis
   (Figure [76]S2C, Supporting Information), suggesting their potential
   roles in immune response regulation. C5_LYVE1 demonstrated typical
   macrophage phagocytic function and showed elevation in the proportion
   within CSF of NMOSD patients (Figure [77]S2C,D, Supporting
   Information).

   During the DEGs analysis of each sub‐cluster within the CSF, a
   significant number of overlapping DEGs were identified among these
   clusters (Figure [78]1E). Among them, these genes predominantly
   included those involved in IFN‐I response, such as IFNAR1, IFNAR2,
   IFITM2, IFITM3, as well as IL17RA, and complement activation genes
   C1QB, C1QC (Figure [79]1F). Subsequently, we performed Gene Set
   Enrichment Analysis (GSEA) for microglia and CNS‐associated
   macrophages‐like cells in the CSF of NMOSD patients and controls. The
   results revealed dysregulated activation of IFN‐I‐related pathways,
   including the “innate immune response”, “type I interferon signaling
   pathway”, and “viral entry into host cells” in the microglia and
   macrophages of NMOSD patients (Figure [80]1G). Given the critical role
   of the cGAS (cyclic GMP‐AMP synthase)‐STING (stimulator of interferon
   genes) pathway in inducing IFN‐I production through the recognition of
   pathogenic DNA nucleic acids, and as a key mechanism for host defense
   against pathogens,^[ [81]^15 , [82]^16 ^] we conducted a comprehensive
   analysis of the cGAS‐STING‐IFN‐I score in CSF microglia and
   macrophages. Our analysis indicated that NMOSD patients exhibited
   significantly elevated cGAS‐STING‐IFN‐I scores in these CSF clusters,
   including C1_HAMP, C3_CCL4, and C5_LYVE1, when compared to the control
   group (Figure [83]1H). The levels of cell free double stranded DNA
   (cfDNA) were directly detected in paired CSF and plasma samples from
   patients with NMOSD and controls. The results indicated that the levels
   of cfDNA in CSF were significantly correlated with those in plasma
   (Figure [84]S2E, Supporting Information), and the levels of cfDNA in
   CSF and plasma of patients with NMOSD were notably higher than those of
   the control group (Figure [85]1I; Figure [86]S2F, Supporting
   Information). Moreover, consistent with DEG analysis results, the
   pathways of “complement activation” and “B cell‐mediated immune
   regulation” were also upregulated in NMOSD patients versus control
   (Figure [87]1G). Collectively, these findings highlight the significant
   role of cGAS‐STING mediated IFN‐I production in the CNS compartment and
   emphasize the involvement of B cells and Th17 cells in the pathogenesis
   of NMOSD.

2.2. cGAS‐STING‐IFN‐I Activation in NMOSD Peripheral Blood Monocytes

   Having identified cGAS‐STING‐IFN‐I activation in CSF myeloid cells, we
   next asked whether similar dysregulation would occur in peripheral
   blood monocytes. We conducted a re‐clustering analysis of peripheral
   blood monocytes and successfully identified six distinct subclusters
   (Figure [88]2A). C1_S100A8, C2_LGALS2, C5_CCL4, and C6_ISG15 were more
   likely to be classified as classical CD14^+ monocytes, while C4_MHC‐II
   were identified as CD14^+CD16^+ intermediate monocytes, and C3_ FCGR3A
   were categorized as CD16^+ non‐classical monocytes (Figure [89]2B).
   Among them, there were no significant differences in the proportions of
   these clusters in peripheral blood, and C3_FCGR3A and C5_CCL4 clusters
   demonstrated heightened pro‐inflammatory functions (Figure [90]2C,D;
   Figure [91]S3B, Supporting Information). Consistent with the
   observations in the CSF, DEGs and GSEA analysis further revealed that
   monocytes subclusters derived from the blood of NMOSD patients also
   exhibited a robust cGAS‐STING‐IFN‐I response (Figure [92]2E,F; Figure
   [93]S3C,D, Supporting Information). Additionally, we analyzed
   cGAS‐STING‐IFN‐I scores for identified main immune cell subtypes and
   found that myeloid cells, particularly, peripheral blood monocytes and
   CSF microglia‐like cells in NMOSD patients exhibited higher
   cGAS‐STING‐IFN‐I scores (Figure [94]S4, Supporting Information). We
   also measured plasma levels of 2′3'‐cGAMP (2′3'‐cyclic GMP‐AMP), a
   crucial component in the cGAS‐STING pathway that has a strong binding
   affinity to STING and potent induction of IFN‐I.^[ [95]^16 ^] Our
   findings revealed a significant increase in the levels of 2′3′‐cGAMP
   within plasma samples in NMOSD patients compared to controls (3.92‐fold
   increase) (Figure [96]2G). These findings underscore the role of
   monocytes and microglia as primary sources of IFN‐I in NMOSD patients
   and implicate cGAS‐STING‐IFN‐I activation in the disease pathogenesis.

Figure 2.

   Figure 2
   [97]Open in a new tab

   Blood monocytes displayed IFN‐I over‐activity similar to that observed
   within the CSF of NMOSD patients. A) UMAP plot depicting the
   subclusters of blood monocytes derived from scRNA‐seq data. B) UMAP
   plot illustrating the expression patterns of canonical markers in blood
   monocytes. C) Dot plot illustrating the differential expression of
   highly expressed genes within distinct subclusters. D) Box plot
   depicting disparities in the proportions of blood monocyte subclusters
   between patients with NMOSD and control subjects. E) UpSet plot
   visualizing the comparative size and overlap of DEGs within blood
   monocyte subclusters between NMOSD patients and controls. F)
   Representative IFN‐I and pro‐inflammatory related genes between NMOSD
   patients and control individuals. G) The plasma levels of 2′3′‐cGAMP
   obtained from both control and patients with NMOSD were quantified by
   ELISA. (control = 21, NMOSD = 36). Data were analyzed using Mann
   Whitney test; Mean ± s.e.m; ***p < 0.001.

2.3. IFN‐I Promotes the Inflammatory Response of T Cells in NMOSD

   Our prior findings suggest the potential involvement of microglia and
   monocytes in the differentiation and maturation process of T cells in
   NMOSD patients. Subsequently, we carried out a further sub‐cluster
   analysis of CD4^+ T cell subsets and defined naive T cells, central
   memory (CM) T cells, Tregs (C7_FOXP3), follicular helper T cells (Tfh),
   Th1 cells, Th2 cells, Th17 cells, and terminal effector memory T cells
   (TEMRA) (Figure [98]3A; Figure [99]S5A–C, Supporting Information).
   Subsequently, we verified the scoring of the defined T cell subtypes,
   thus confirming the classification of Th1, Th2, and Th17 subtypes
   (Figure [100]3B). Notably, our results showed that the CSF and
   peripheral blood of NMOSD patients had higher scores for CD4^+ T cell
   subtypes (Figure [101]3C), along with more robust cellular activation
   and cytotoxic scores (Figure [102]S5D,E, Supporting Information). These
   findings indicated a stronger inflammatory response in CD4^+ T cells of
   NMOSD patients. We next explored the effect of IFN‐I on the
   differentiation of CD4^+ T cells by pseudotime analysis. Initially,
   CD4^+ T cells manifested several distinct major differentiation
   trajectories: from the naive state into the Th1 cell type and
   subsequently evolved into terminal effector memory T cells; from the
   naive state into the Th17 cell subtype; from the naive state to the
   follicular helper T cell state, and from the naive state to the Treg
   cell state (Figure [103]3D,E). Among these, CD4^+ T cells derived from
   patients with NMOSD demonstrated a higher intensity of differentiation
   in the middle and late stages of the differentiation process
   (Figure [104]3F). Subsequently, we found that the expression of IFN‐I
   response‐related genes showed a gradually increasing trend during the
   differentiation of CD4^+ T cells from the naive state to the terminally
   effector state (Figure [105]3G). Specifically, in the Th1
   differentiation trajectory, IFN‐I response‐related genes such as IRF1,
   STAT1, and IFI16 showed a significant up‐regulation in the expression
   from the naive state to the terminal state (Figure [106]3H). In the
   Th17 differentiation trajectory, the expression of IFN‐I
   response‐related genes such as IFI16, IFITM2, ISG15, and GBP5 gradually
   increased (Figure [107]3H). In addition, during the differentiation
   process, the IFN‐I response‐related pathway score of NMOSD patients was
   significantly higher than that of the control group (Figure [108]3I).
   These findings collectively demonstrate an enhanced inflammatory
   response in CD4^+ T cells from NMOSD patients, where elevated IFN‐I
   activity further amplified the inflammatory responses of CD4^+ T cells,
   potentially contributing to disease pathogenesis.

Figure 3.

   Figure 3
   [109]Open in a new tab

   IFN‐I facilitated the development of CD4^+ T cells in NMOSD. A) UMAP
   plot depicting the subclusters of CD4^+ T cells derived from all
   samples of scRNA‐seq data. B) The gene‐set scores of different T helper
   subsets (Th1, Th2, and Th17 gene signatures) in CD4^+ T cells. C) The
   gene scores of different helper T cell subsets (characterized by Th1,
   Th2, and Th17 gene signatures) were compared between the control group
   and patients with NMOSD. D,E) UMAP plot showing pseudotime trajectory
   of CD4^+ T cells based on Slingshot demonstrated three distinct
   trajectories (Th1, Th17, and Tfh). F) The density plot of CD4^+ T cell
   differentiation trajectories showing the changes in the intensity of
   different T cell subtypes during the differentiation process between
   the NMOSD group and the control group. G) Heat map illustrating the
   normalized log count of gene expression associated with IFN‐I response
   in CD4^+ T cells during the trajectory of differentiation and
   maturation. H) The expression levels of representative genes driving
   CD4^+ T cells development along Th1 and Th17 trajectories. I) The
   differences in IFN‐I‐related pathway scores between the NMOSD group and
   the control group during the differentiation process of CD4^+ T cells.

2.4. Activation of cGAS‐STING‐IFN‐I Pathway in NMOSD Mouse Models

   To further confirm the aberrant activation of the cGAS‐STING‐IFN‐I
   pathway in microglia‐like cells identified in the CSF of NMOSD
   patients, we first utilized the NMO‐IgG intracerebral injection mouse
   model (NMO‐IgG model)^[ [110]^17 ^] which is capable of mimicking the
   response of microglia subsequent to the direct attack of NMO‐IgG on
   astrocytes in the CNS during the pathogenesis of NMOSD. We performed
   bulk RNA‐sequencing (RNA‐seq) analysis on fluorescence‐activated cell
   sorted CD45^int CD11b^+ microglia from wild‐type (WT) mice, control‐IgG
   injected mice, and NMO‐IgG injected mice (Figure [111]4A).
   Transcriptional analysis revealed that mice in the NMO‐IgG injected
   group showed an upregulation of cGAS‐STING‐IFN‐I pathway related genes
   expression in microglia, including Cgas, Isg15, and Zbp1. This was
   accompanied by the activation of disease‐associated microglia (DAM)
   genes^[ [112]^18 ^] and an increase in chemokine‐related genes
   expression (Figure [113]4B). Meanwhile, qPCR experiments conducted on
   CD11b^+ microglia isolated by magnetic cell sorting (MACS) from brain
   tissue revealed that NMO‐IgG injection resulted in increased expression
   of genes associated with the cGAS‐STING‐IFN‐I pathway, as compared to
   WT and control‐IgG injected mice. This included increased expression of
   Tmem173 gene (encoding STING), Irf3, Ifna1, Ifnar1, Cxcl10, and Isg15
   at the peak of the disease. Activation of microglia activation genes,
   including Cd68, Il‐1b, Clec7a, and C1qa were also upregulated
   (Figure [114]4C). As mitochondrial DNA (mtDNA) leakage is also a core
   activator of the cGAS‐STING signaling pathway,^[ [115]^19 ^] we next
   investigated the changes in mitochondrial homeostasis in NMOSD mouse
   models using transmission electron microscopy (TEM). We found that the
   mitochondria in the microglia of mice treated with NMO‐IgG exhibited
   abnormal morphology, identified by the loss of cristae and/or the
   disruption of the mitochondrial outer and/or inner membrane
   (Figure [116]4D,E). Immunofluorescence analysis also revealed an
   increase in DNA accumulation within IBA1^+ microglia in the NMO‐IgG
   group (Figure [117]4F,G). The key protein IRF3 in the cGAS‐STING‐IFN‐I
   pathway also showed increased phosphorylation levels in IBA1^+
   microglia from NMO‐IgG group (Figure [118]S6A,B, Supporting
   Information).

Figure 4.

   Figure 4
   [119]Open in a new tab

   Enhanced cGAS‐STING‐IFN‐I activity in NMOSD Mouse Models. A) Schematic
   representation of NMO‐IgG injection model. B) Heatmap illustrating the
   expression levels of cGAS‐STING‐IFN‐I, disease‐associated microglia
   (DAM), and chemokine genes derived from bulk RNA‐seq of microglia
   sorted from WT mice, Control‐IgG treated mice, and NMO‐IgG treated mice
   (n=3 per group). C) Quantitative PCR analysis revealed altered
   expression for cGAS‐STING‐IFN‐I and microglia activation‐related genes
   in MACS‐sorted CD11b^+ microglia from WT mice, Control‐IgG‐treated
   mice, and NMO‐IgG‐treated mice during the peak of disease (n = 5 per
   group). D,E) Representative transmission electron microscopy images and
   quantitative analysis of damaged mitochondrial alterations in microglia
   of control‐IgG‐treated mice and NMO‐IgG‐treated mice (n = 4 per group).
   F,G) Representative immunofluorescence staining images and quantitative
   analysis of the changes in DNA mean fluorescence intensity within
   IBA1^+ microglia in WT mice, control‐IgG‐treated mice, and
   NMO‐IgG‐treated mice (n = 4 per group). H) Flow chart depicting the
   acquisition and in vitro culture procedure of AQP4‐specific T cells. I)
   Representative flow cytometry and statistical analysis demonstrated
   that IFN‐I promoted an increase in the proportion of
   I‐A^b‐AQP4(205‐215) tetramer‐positive T cells within CD4^+ T cells
   (n = 6–8). J) Representative flow cytometry and statistical analysis
   revealed that under Th17 polarization conditions, IFN‐I promoted the
   increase in the proportion of IFN‐γ and IL‐17A positive T cells in
   CD4^+ T cells (n = 6). K) Representative flow cytometry and statistical
   analysis demonstrated that IFN‐I promoted an increase in the proportion
   of chemokine receptors CCR5, CCR6, and CXCR3 positive T cells in CD4^+
   T cells (n = 6). Data presented as Mean ± s.e.m; *p < 0.05, **p < 0.01,
   ***p < 0.001, P values were calculated by the one‐way ANOVA test,
   Kruskal‐Wallis test, or Mann‐Whitney test.

   To investigate the effect of IFN‐I response on AQP4‐specific T cells,
   we next immunized Aqp4 ^‐/‐ mice with the previously reported
   AQP4‐specific peptides AQP4[p135‐153] or AQP4[p201‐220] to obtain
   AQP4‐specific T cells for further study.^[ [120]^20 , [121]^21 ^]
   First, we immunized Aqp4 ^‐/‐ mice with the AQP4[p135‐153] peptide, and
   then obtained lymphocytes from draining lymph nodes for Th17
   polarization (IL‐23, IL‐6, and the AQP4‐specific peptide), followed by
   passive transfer (Figure [122]S6C, Supporting Information).^[ [123]^20
   ^] In the Th17‐AQP4[p135‐153] passive transfer mice, we further
   validated the changes of the cGAS‐STING‐IFN‐I pathway in MACS sorted
   microglia from spinal cord tissue during the peak of disease at 14 days
   (Figure [124]S6D, Supporting Information). Next, we immunized Aqp4^‐/‐
   mice with the AQP4[p201‐220] peptide. Lymphocytes were obtained from
   the draining lymph nodes and cultured in vitro (Figure [125]4H). To
   directly explore the impact of IFN‐I on AQP4‐specific T cells, we
   employed the I‐A^b‐AQP4(205‐215) tetramers.^[ [126]^21 ^] The results
   indicated that the additional administration of IFN‐I would elevate the
   proportion of CD4^+ AQP4‐specific autoreactive T cells
   (Figure [127]4I). Moreover, we observed that, in comparison with the
   control group and the Th17 polarization condition, additional IFN‐I
   stimulation could increase the proportion of IFN‐γ and IL‐17A positive
   cells within CD4^+ T cells (Figure [128]4J). Simultaneously, Th17
   polarization condition would also induce an increase in the proportion
   of CCR5^+ CD4^+ and CCR6^+ CD4^+ T cells (Figure [129]4K). Moreover,
   when additional IFN‐I stimulation was used under Th17 polarization
   conditions, this would significantly result in an increase in the
   proportion of CCR5^+ CD4^+, CCR6^+ CD4^+, and CXCR3^+ CD4^+ T cells
   (Figure [130]4K).

2.5. Sting Deficiency Alleviated NMO‐IgG Associated Pathology in NMOSD

   We next developed Sting gene knockout mice and established NMO‐IgG
   model to investigate the role of cGAS‐STING‐IFN‐I in the pathogenesis
   of NMOSD (Figure [131]5A). We first performed immunofluorescence
   analysis, defining cells co‐expressing IBA1 and C1q as activated
   microglia, and cells co‐expressing GFAP and C3 as activated
   astrocytes.^[ [132]^4 , [133]^18 ^] Our results revealed a significant
   reduction in the proportion of activated glial cells surrounding the
   lesion site in Sting^‐/‐ mice compared to Sting^+/+ mice (microglia:
   38.9% vs 16.2%; astrocytes: 16.6% vs 7.4%) (Figure [134]5B,C). In
   addition, reduced levels of IRF3 and TBK1 phosphorylation were observed
   in IBA1^+ microglia surrounding the lesion site in Sting^‐/‐ mice
   (Figure [135]5D,E). Moreover, using 9.4‐T MRI to visualize
   demyelination lesions (T2) on day 5, a significant reduction in the
   volume of demyelination lesions was observed in Sting^‐/‐ mice (17.1 vs
   9.4 mm^3) (Figure [136]5F,G). These findings suggest that inhibition of
   STING effectively suppressed the excessive activation of the IFN‐I
   response mediated by NMO‐IgG, which directly attacked astrocytes in
   NMOSD.

Figure 5.

   Figure 5
   [137]Open in a new tab

   Sting^‐/‐ mice exhibited improved immunopathology in the NMO‐IgG model.
   A) Flow chart illustrating the process of NMO‐IgG intracerebral model.
   B) Representative immunofluorescence images and quantification of the
   IBA1^+/C1q^+ microglia around the lesion in the brain of WT mice and
   Sting^‐/‐ mice at the peak of disease (n = 5 per group). C)
   Representative immunofluorescence images and quantitative analysis
   showing changes in GFAP^+/C3^+ astrocytes around the lesion in
   Sting^‐/‐ compared to WT mice at the peak of disease (n = 5 per group).
   D) Representative immunofluorescence images and quantitative analysis
   of phosphorylated IRF3 (pIRF3) mean fluorescence intensity (MFI) around
   the lesions in WT and Sting^‐/‐ mice at the peak of disease (n = 5 per
   group). E) Representative immunofluorescence images and quantitative
   analysis of phosphorylated TBK1 (pTBK1) MFI around the lesions in WT
   and Sting^‐/‐ mice at the peak of disease (n = 5 per group). F)
   Representative MRI images showing lesion volume at the peak of the
   disease between WT and Sting^‐/‐ mice. G) MRI analysis quantifies the
   demyelinating lesion volume in both WT and Sting^‐/‐ mice at the peak
   of disease (n = 5 per group). Data were analyzed using unpaired
   student's t‐test or Mann‐Whitney test; Mean ± s.e.m; *p < 0.05,
   **p < 0.01.

2.6. Reduction of AQP4‐Specific T Cell Inflammation and Autoreactivity by
Sting Deficiency

   To further investigate the role of IFN‐I in the pathogenesis of NMOSD
   on AQP4‐specific T cells, we then used AQP4‐specific passive transfer T
   cell model to study the effect of the STING‐IFN‐I pathway (Figure
   [138]6A). First, in the mouse model of Th17 AQP4[p135‐153] adoptive
   transfer, we discovered Sting^‐/‐ mice exhibited faster symptom
   recovery with lower clinical scores in comparison to the WT (maximal
   mean clinical score: 2.4 vs 1.1) (Figure [139]6B). Histological
   analyses, including H&E staining and LFB staining, showed reduced
   inflammatory cell infiltration and demyelination in spinal cord lesions
   in the Sting^‐/‐ mice (Figure [140]6C,D). At peak disease, Sting^‐/‐
   mice had less activated TEME119^+/C1q^+ microglia (32.8% vs 16.3%) and
   GFAP^+/C3^+ astrocytes (25.82% vs 10.72%) in their spinal cord lesions
   compared to WT mice (Figure [141]6E,F). Additionally, the
   phosphorylation levels of TBK1 and IRF3 were reduced in IBA1^+
   microglia within spinal cord lesions of Sting^‐/‐ mice
   (Figure [142]6G,H). Knockout of Sting also reduced IFN‐I‐associated
   genes in MACS‐sorted CD11b^+ microglia during the peak disease
   (Figure [143]6I).

Figure 6.

   Figure 6
   [144]Open in a new tab

   Attenuated glia damage and IFN‐I activation in AQP4 specific Sting^‐/‐
   deficiency mouse model. A) Schematic representation of Th17‐AQP4
   specific T cells passive transfer model. B) Clinical scores of the
   Th17‐AQP4 model were compared between WT and Sting^‐/‐ mice (n = 8 per
   group). C) Representative H&E and LFB staining images of Th17‐AQP4
   model between WT and Sting^‐/‐ mice, scale=500 µm. D) Quantitative
   analysis revealed a significant decrease in inflammatory infiltration
   and demyelination level among Sting^‐/‐ mice (n = 6 per group). E)
   Representative immunofluorescence images and quantitative analysis were
   conducted to assess the presence of TMEM119^+/C1q^+ microglia of WT
   mice and Sting^‐/‐ mice within spinal cord lesions at the peak of the
   disease, indicating a significant reduction in microglial activation in
   Sting^‐/‐ mice compared to WT mice (n = 6 per group). F) Representative
   immunofluorescence images and quantitative analysis of the GFAP^+/C3^+
   activated astrocytes in spinal cord lesions, showing lower levels of
   astrocyte activation in Sting^‐/‐ mice during the peak of the disease
   (n = 5 per group). G) Representative immunofluorescence images and
   quantitative analysis of pTBK1 mean fluorescence intensity were
   compared between IBA1^+ microglia within spinal cord lesions of
   Sting^‐/‐ mice and WT mice (n = 5 per group). H) Representative
   immunofluorescence images and quantitative analysis of the MFI of pIRF3
   were compared between IBA1^+ microglia within the spinal cord lesions
   of Sting^‐/‐ mice and WT mice (n = 5 per group). I) qPCR analysis
   showed decreased expression of Ifna1, Ifnar1, Irf3, Cxcl10, Il‐1b, and
   C1qa genes in the microglia of Sting^‐/‐ mice during the peak of
   disease (n = 5 per group). Statistical analysis for clinical scores was
   performed using one‐way ANOVA with Tukey's post‐test, while unpaired
   student's t‐test or Mann‐Whitney test was employed for histology and
   flow cytometry comparisons; data presented as Mean ± s.e.m.; *p < 0.05,
   **p < 0.01, ***p < 0.001.

   Meanwhile, in the AQP4[p135‐153] specific T cell adoptive transfer
   model, flow cytometry analysis revealed significant reductions in CD4^+
   T cells, CD8^+ T cells, B220^+ B cells, and Ly6G^+ neutrophils (3.92‐,
   3.39‐, 2.63‐, and 2.60‐ fold decrease, respectively) in the CNS during
   disease peak in Sting^‐/‐ mice. The population of CD45^intCD11b^+
   microglia exhibited a 1.33‐fold decrease, accompanied by a reduction in
   the proportion of CD86^+ cells within these microglia (Figure
   [145]7A,B). Meanwhile, the proportions of IL‐17A‐positive cells in
   CD4^+ T cells (11.03% vs 5.1%) and IL‐6‐positive cells within B cells
   were both significantly decreased (7.46% vs1.94%) (Figure [146]7C,D).
   While the relative proportions of T cell, B cell, and other immune cell
   populations remained unchanged in both the spleen and blood, there was
   also a significant decrease in the proportion of IL‐17A‐positive CD4^+
   T cells (0.79% compared to 0.55%) and IL‐6‐positive cells in B cells
   (1.07% compared to 0.61%) within the spleen (Figure [147]S7A–D,
   Supporting Information). No significant changes in the proportion of T
   and B cells within the CNS of WT and Sting^‐/‐ mice were observed
   during the disease recovery phase (Figure [148]S7E, Supporting
   Information). The co‐culture of AQP4[p135‐153]‐specific CD4^+ T cells
   and microglia, sorted by FACS, also revealed that Sting deficiency
   resulted in a significant reduction in the proportion of IL‐17A
   positive cells (6.25% vs 4.13%) and IFN‐γ‐positive cells (4.80% vs
   3.50%) within the CD4^+ T cell population (Figure [149]S7F–H,
   Supporting Information). The CFSE assay in vitro demonstrated that
   Sting deficiency in microglia and monocytes could effectively suppress
   the expansion of AQP4 [p135‐153]‐specific CD4^+ T cells
   (Figure [150]7E,G). Following the adoptive transfer of
   AQP4[p201‐220]‐specific T cells, antigen‐specific T cells were detected
   in both the WT mice and Sting^‐/‐ mice. The results indicated that
   Sting^‐/‐ mice presented a decreased proportion of I‐A^b‐AQP4(205‐215)
   tetramer positive T cells (Figure [151]7H). Meanwhile, the proportion
   of chemokine receptor CCR5^+, CCR6^+, and CXCR3^+ cells within CD4^+ T
   cells in the CNS of Sting^‐/‐ mice also decreased significantly
   (Figure [152]7I). Sting was knocking down in AQP4[p201‐220] specific T
   cells by using retrovirus encoding shRNA. We discovered that under Th17
   polarization conditions in vitro, the knockdown of Sting could
   significantly reduce the proportion of IL‐17A and IFN‐γ positive cells
   within CD4^+ T cells (Figure [153]S8A,B, Supporting Information).
   Following adoptive transfer of retrovirus‐transduced cells into WT
   mice, compared to those receiving shRNA‐NC transduced cells, mice
   receiving shRNA‐Sting transduced cells exhibited a significant
   reduction in the proportions of CD4⁺ T cells, CD8⁺ T cells, and B220⁺ B
   cells infiltrating the CNS (Figure [154]S8C, Supporting Information).
   Meanwhile, the proportion of I‐A^b‐AQP4(205‐215) tetramer positive T
   cells also significantly decreased (Figure [155]S8C, Supporting
   Information). These findings indicate that STING inhibition could
   effectively attenuate disease severity and reduce activation of
   AQP4‐specific T cells.

Figure 7.

   Figure 7
   [156]Open in a new tab

   Sting^‐/‐ deficiency hindered the activation of AQP4‐specific T cells.
   A) Flow cytometry gating strategy of different cell types in the CNS of
   AQP4[p135‐153] specific T cells transfer mice. B) Quantitative analysis
   showed a notable reduction in the proportion of the infiltrating immune
   cells within the CNS of WT and Sting^‐/‐ mice (n = 8 per group). C,D)
   The frequency of IL‐17A^+ cells within the CD4^+ T cells and the
   percentage of IL‐6^+ cells among B cells were assessed using flow
   cytometry in both WT and Sting‐/‐ mice (n = 8 per group). E) Flow chart
   depicting the process of CFSE assay in vitro. F,G) Representative flow
   cytometry and statistical analysis indicated that the absence of Sting
   in CNS microglia and periphery monocytes reduced the expansion of
   AQP4‐specific T cells in vitro (n = 5–8 per group). H) Representative
   flow cytometry and statistical analysis indicated that following the
   adoptive transfer of AQP4[p201‐220] T cells, the proportion of
   antigen‐specific T cells in the Sting^‐/‐ mice was lower compared to WT
   mice (n = 5 per group). I) Flow cytometry results were analyzed to
   observe the alterations in the proportion of CD4^+ chemokine
   receptor‐positive T cells within the CNS tissues of WT and Sting^‐/‐
   mice following the adoptive transfer of AQP4[p201‐220] T cells. The
   gating strategy was consistent with that in Figure [157]4K (n = 5 per
   group). Unpaired student's t‐test or Mann‐Whitney test was employed to
   compare results across groups; data presented as Mean ± s.e.m;
   *p < 0.05, **p < 0.01, ***p < 0.001.

2.7. Treatment with STING Inhibitor H‐151 Alleviated NMOSD Mouse Models

   To further validate the therapeutic targeting of the cGAS‐STING‐IFN‐I
   pathway, we employed a specific STING inhibitor, H‐151,^[ [158]^22 ^]
   to treat NMOSD in mouse models. In the Th17‐AQP4[p135‐153] passive
   transfer model, mice receiving H‐151 treatment exhibited notable
   improvements in clinical symptom scores compared to the control group
   (maximal mean clinical score: 2.5 vs 1.9) (Figure [159]8A,B). At the
   disease peak, immunofluorescence staining of spinal cord lesions
   revealed a remarkable reduction in both microglial and astrocyte
   activation in H‐151 treated mice compared to WT mice
   (Figure [160]8C–F). Flow cytometry analysis further confirmed a
   significant reduction in the infiltration of inflammatory cells,
   including CD4^+ T cells, CD8^+ T cells, B220^+ B cells, and Ly6G^+
   neutrophils (fold decreases of 7.12, 6.01, 4.38, and 3.90,
   respectively) in mice treated with H‐151 (Figure [161]8G). It is
   noteworthy that, while the overall number of CD45^intCD11b^+ microglia
   remained constant in the CNS, there was a notable reduction in the
   percentage of CD86‐positive cells within CD45^intCD11b^+ microglia
   (Figure [162]8G). Additionally, in the NMO‐IgG model, mice treated with
   H‐151 also demonstrated a significant reduction in microglia and
   astrocyte activation compared to WT mice (Figure [163]S9, Supporting
   Information). These findings collectively support the potential of
   targeting the STING pathway as a therapeutic approach for NMOSD
   treatment.

Figure 8.

   Figure 8
   [164]Open in a new tab

   STING inhibitor H‐151 effectively improved the AQP4‐specific T cells
   model. A) Schematic representation of Th17‐AQP4 model treated with
   STING inhibitor H‐151. B) Clinical symptom scores compared between
   control group mice and mice treated with H‐151 in the Th17 mouse model
   (n = 8 per group). C) Representative immunofluorescence images
   depicting changes in TMEM119^+/C1q^+ microglia within the spinal cord
   lesion of WT mice and H‐151‐treated mice at the peak of disease. D)
   Quantitative analysis of TMEM119^+/C1q^+ microglia (n = 5 per group).
   E) Representative immunofluorescence images demonstrating alterations
   in GFAP^+/C3^+ astrocytes within the spinal cord lesion of WT mice and
   H‐151‐treated mice at the peak of disease. F) Quantitative analysis
   revealing changes in GFAP^+/C3^+ astrocyte (n = 5 per group). G) Flow
   cytometry analysis to quantify CNS‐infiltrated cells in both WT and
   H‐151‐treated mice at the peak of disease. The gating strategy was
   consistent with that in Figure [165]7A (n = 5 per group). Statistical
   analyses: Clinical scores were assessed using one‐way ANOVA with
   Tukey's post‐test. Unpaired student's t‐test or Mann‐Whitney test was
   used to compare results from flow cytometry and immunofluorescence
   studies. Data were presented as Mean ± s.e.m, *p < 0.05, **p < 0.01,
   ***p < 0.001.

3. Discussion

   Recent studies imply that a dysregulated recognition of self‐DNA
   through the cGAS‐STING pathway contributes to the development of
   autoimmune disorders.^[ [166]^23 , [167]^24 , [168]^25 , [169]^26 ^]
   Whether and how this pathway operates in CNS autoimmune diseases, i.e.,
   NMOSD, remains unexplored. Previous studies have demonstrated that
   cfDNA derived from neutrophil extracellular trap‐related cell death
   (NETosis) can induce abnormal IFN‐I responses in the peripheral blood
   of patients with NMOSD.^[ [170]^27 ^] In our study, not only was an
   increase in cfDNA concentration detected in the plasma of NMOSD
   patients, but also in their CSF. Simultaneously, the leakage of
   mitochondrial DNA is a crucial factor in activating the cGAS‐STING.
   Previous studies have shown that during aging, the leakage of
   mitochondrial DNA in microglia leads to the activation of the
   cGAS‐STING pathway.^[ [171]^19 ^] Our findings also indicate that
   NMO‐IgG could cause abnormal mitochondrial damage and abnormal
   intracellular DNA accumulation in microglia. These results suggest that
   in NMOSD, extracellular cfDNA and intracellular mitochondrial DNA
   leakage in the CNS might jointly contribute to the activation of the
   cGAS‐STING pathway in microglia. After cGAS activation, cGAMP is
   usually released and absorbed by bystander cells via gap junctions,
   resulting in the activation of STING beyond the actual site of cGAS
   activation.^[ [172]^15 , [173]^16 ^] Consequently, elevated plasma
   cGAMP levels in patients with NMOSD suggest that it may function by
   STING‐dependent IFN‐I production.

   In our scRNA‐seq subcluster analysis of myeloid cells in both the CSF
   and peripheral blood, we identified a specific cluster exhibiting
   expression of the chemokine CCL4. These CCL4^+ microglia‐like cells and
   monocytes may regulate Th17 cell differentiation, as evidenced by the
   accompanying upregulation of IL17RA in NMOSD. The GSEA analysis
   identified the activation of B cell‐mediated immunity in the microglia
   and macrophages in the CSF, further substantiating the role of IFN‐I
   response in driving aberrant activation in NMOSD patients. Further,
   this analysis reinforces the regulatory role of IFN‐I in modulating the
   inflammatory responses of T cells and B cells in NMOSD patients.^[
   [174]^2 , [175]^28 , [176]^29 , [177]^30 ^] Our analysis of CD4^+ T
   cell subclusters in NMOSD patients reveals an augmentation of T cell
   inflammatory response. Pseudotime analysis indicated an elevated
   expression of IFN‐I response genes during T‐cell differentiation and
   maturation, underscoring the significant impact of IFN‐I on T cells
   response in the NMOSD.^[ [178]^8 , [179]^31 ^] A previous study has
   shown that the frequency of AQP4‐specific effector memory CD4^+ T cells
   in patients with NMOSD were increased with an effector memory
   phenotype, and the autoreactive T cells in patients with NMOSD may
   undergo an exhaustion‐like phenotype under chronic stimulation. This
   phenotype was characterized by a gradual loss of pro‐inflammatory
   cytokine expression and acquisition of co‐inhibitory receptors, along
   with the expression of FOXP3. Although these cells show exhaustion
   features, they still retain the ability of B cell helper function.^[
   [180]^32 ^] In our study, IFN‐I pathway related genes showed increased
   expression when T cells were differentiated into the terminal stage.
   During in vitro culture of AQP4‐specific T cells, we found that IFN‐I
   increased the proportion of AQP4 antigen‐specific T cells. However, the
   AQP4‐specific T cells of patients with NMOSD did not show an abnormal
   IFN‐I response in a previous study.^[ [181]^32 ^] We speculate that the
   difference may be due to the fact that some of the patients in that
   study were in the period of remission of NMOSD. In contrast, all the
   enrolled patients in our study were in the acute phase. During the
   disease active stage, IFN‐I promotes the presentation of AQP4 antigen
   by antigen‐presenting cells by enhancing the expression of MHC‐II
   molecules,^[ [182]^33 ^] accelerating the activation and clonal
   expansion of autoreactive T cells, which may be a key mechanism in the
   acute stage of the disease. However, long‐term exposure to IFN‐I may
   induce the expression of co‐inhibitory receptors (such as TIM‐3 and
   LAG‐3) in T cells. This dynamic balance may manifest as the
   “self‐limiting” property of IFN‐I signaling.^[ [183]^34 ^] Meanwhile,
   we found that the additional use of IFN‐I under the Th17 polarization
   conditions (IL‐6 and IL‐23) could simultaneously increase the
   proportion of IL‐17A and IFN‐γ positive T cells. This finding
   reminiscent a previous study which also showed that in the Th17‐EAE
   model, the additional treatment with IFN‐β leads to an increase in
   GM‐CSF and IL‐17 positive T cells in mice.^[ [184]^8 ^] Given that the
   treatment of IFN‐I has an immunomodulatory effect of inhibiting Th1
   cells in MS,^[ [185]^35 ^] our study further reveals the dual role of
   IFN‐I in regulating T cell responses under complex autoimmune states.

   Here, we provide molecular evidence suggesting that intrinsic
   activation of the myeloid cell IFN‐I response is implicated in NMOSD
   exacerbation. In contrast to the worsening effects of IFNβ‐1b, blocking
   the cGAS‐STING‐IFN‐I pathway can mitigate NMO‐IgG‐induced CNS
   pathogenicity and reduce the activation of AQP4‐specific T cells. Our
   research demonstrated that Sting deficiency significantly improved
   clinical symptoms in NMOSD mouse models by reducing the activation of
   microglia and astrocytes, as well as AQP4‐specific T cells. This is
   consistent with previous studies that demonstrated a correlation
   between aberrant activation of the IFN‐I pathway and microglia in mouse
   models and Ifnar knockout alleviated pathological changes in NMOSD.^[
   [186]^36 ^] Furthermore, a recent study using AQP4[p201‐220] induced
   mouse model showed that Ifnar ^‐/‐ mice had a lower incidence and
   clinical score.^[ [187]^37 ^]

   Anifrolumab, a monoclonal antibody targeting IFNAR, has been approved
   for treating systemic lupus erythematosus due to its effective
   inhibition of the IFN‐I activity.^[ [188]^38 , [189]^39 ^] Currently, a
   range of specific pharmaceuticals targeting cGAS or STING are also
   under development as potential clinical candidates for the treatment of
   autoimmune disorders.^[ [190]^40 ^] For instance, H‐151, serving as an
   effective STING inhibitor in both human and murine models, exhibits
   significant suppression of systemic cytokine response through reduction
   of IFN‐I reaction.^[ [191]^22 ^] Hence, for NMOSD patients with
   aberrant activation of the cGAS‐STING‐IFN‐I pathway, targeting this
   pathway may offer novel therapeutic options that could concurrently
   influence myeloid cells and self‐reactive T and B cells. These findings
   warrant further investigation to assess the potential of such therapies
   in the treatment landscape of NMOSD.

4. Experimental Section

Process of scRNA‐Seq Sample Acquisition

   Five AQP4‐IgG seropositive NMOSD patients were enrolled from Tianjin
   Medical University General Hospital. All participants fulfilled the
   2015 criteria for NMOSD as established by the International Panel for
   NMO Diagnosis (IPND).^[ [192]^41 ^] Clinical information was obtained
   from the medical records of hospitalized patients. Before
   participation, all individuals provided informed consent. Full informed
   consent was acquired and the work was carried out under the approvals
   of the Research ethics committee IRB2021‐KY‐134. Additionally, paired
   CSF and PBMCs samples were collected from 3 healthy controls and 5
   control non‐neuroinflammatory samples from the GEO database
   ([193]GSE133028, [194]GSE138266) to serve as control groups for
   subsequent analysis. The demographic characteristics of patients and
   controls are detailed in Table [195]S1 (Supporting Information). For
   each newly enrolled patient with NMOSD, 5 ml of peripheral blood and
   10 ml of CSF samples were collected for subsequent experimental
   analysis. Whole blood and CSF samples were processed similarly to the
   previously described methodology.^[ [196]^9 , [197]^10 , [198]^11 ^]
   CSF cells were obtained via centrifugation, and PBMCs were isolated
   from fresh whole blood using the Ficoll‐Paque density gradient method.
   One CSF sample could not be used for subsequent experiments due to low
   cell activity. For scRNA‐seq library preparation, 5′ library
   preparation kits from 10x Genomics (Pleasanton, CA, USA) were utilized.

scRNA‐seq Data Analysis

   Due to the incorporation of scRNA‐seq samples from various studies into
   the integrated dataset, data quality control was initially conducted on
   the expression matrices of all included samples. In order to address
   batch effects and harmonize the data from all samples, the canonical
   correlation analysis (CCA) algorithm in Seurat (v4.3.0.1) was employed
   to mitigate confounding factors such as different sequencing platforms
   and tissues. The typical UMAP plots before and after data integration
   are presented in Figure [199]S1B (Supporting Information). Cell
   clustering was performed with the graph‐based clustering algorithm in
   the FindClusters function, set at a resolution of 0.6. UMAP
   visualization was used to display the identified clusters. The Wilcoxon
   test was utilized for differential gene analysis using the
   FindAllMarkers function. Additionally, covariates such as age and
   gender were incorporated into the analysis process to mitigate batch
   effects, with a log fold change threshold set at 0.25. Enriched
   pathways were evaluated through hypergeometric testing in both Gene
   Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   databases. The compareCluster function in the clusterProfiler package
   was used for pathway enrichment analysis. Pathways with a
   Benjamini–Hochberg corrected p‐value < 0.05 were considered
   significantly enriched. To illustrate possible differential activity
   among cell groups and disease conditions, Gene Set Enrichment Analysis
   (GSEA) score calculation was performed for monocyte and microglia‐like
   cell sub‐clusters. Hallmark gene sets were obtained from MSigDB,
   excluding overlapping genes within each gene set. The AUCell package
   (V1.12.0)^[ [200]^42 ^] was used to calculate the signature score of a
   specific gene set. First the ranked expression matrix was built using
   the AUCell_buildRankings function, and then calculated the AUC value
   using the AUCell_calcAUC function. The genes in the utilized gene sets
   are shown in Table [201]S4 (Supporting Information). The cGAS‐STING‐
   IFN‐I gene set was obtained from the relevant pathway genes in the
   MSigDB database. Gene sets associated with Th1, Th2, Th17, T cell
   activation, proliferation, and cytotoxicity, were compiled based on
   previous research.^[ [202]^43 ^] The Monocle package (V2.18.0) was used
   to reconstruct the cellular differentiation trajectory of CD4^+ T cell
   subsets. The naive T cell cluster (C1_CCR7) was set as roots of the
   trajectory to order the cells in pseudotime by using orderCells
   function. The plot_pseudotime_heatmap and plot_genes_in_pseudotime
   function were used to visualize the dynamic expression changes of
   significantly DEGs identified by the differentialGeneTest function
   across the pseudotime of each trajectory. Subsequently, the
   AUCell_calcAUC function was employed to calculate gene‐set scores along
   each identified trajectory. To confirm the trajectory results from
   Monocle, Slingshot (V2.10.0) was applied to infer the differentiation
   trajectory.

CSF and Plasma cfDNA Analysis

   Cell‐free double strand DNA was detected in paired CSF and plasma
   samples from 20 NMOSD patients in the acute phase (first diagnosed or
   relapse), and 12 age and sex matched control subjects. cfDNA was
   analyzed using the Quant‐iT™ dsDNA (Thermo Scientific) in accordance
   with manufacturer's instructions.

Plasma 2′3′‐cGAMP Detection by ELISA

   Plasma samples were obtained from 36 patients diagnosed with NMOSD, as
   well as from 21 control subjects, for the purpose of measuring the
   levels of 2′‐3′ cGAMP. Among them, 32 samples were from the same
   subjects as those used for the above‐mentioned detection of plasma
   cfDNA. The concentration of 2′‐3′ cGAMP was determined using
   enzyme‐linked immunosorbent assays (ELISA), per the manufacturer's
   instructions (Invitrogen, EIAGAMP).

Animals

   Sting^‐/‐ mice (Strain NO. T012747) and Aqp4^‐/‐ mice (Strain NO.
   T011901) were obtained from GemPharmatech (Nanjing, China). Wild‐type
   female C57BL/6 mice, aged 8‐10 weeks old, were purchased from Vital
   River (Beijing, China). All the procedures of animal experiments had
   received approval from the Animal Experimental Ethics Committee of
   Tianjin Medical University (IRB2021‐DWFL‐216) and were conducted
   following the Revised Guide for the Care and Use of Laboratory Animals.

NMO‐IgG Preparation and Intracerebral Injection Model Induction

   The intracerebral NMO‐IgG injection mice model was established
   following methodologies from previous studies.^[ [203]^17 ^] The
   isolation of NMO‐IgG from NMO patients’ serum followed a previously
   described methodology. Both NMO‐IgG and Control human‐IgG (Solarbio,
   Beijing, China) were in a 10 mg ml^−1 concentration in
   phosphate‐buffered saline (PBS). Mice were anesthetized and mounted
   onto a stereotactic frame (RWD Life Science, Shenzhen, Guangdong,
   China). A burr hole was drilled in the skull 2.5 mm to the right of the
   bregma under a midline scalp incision. A 30‐gauge needle attached to a
   50 ul gas‐tight glass syringe (Hamilton, Reno, NV, USA) was inserted
   3 mm deep to infuse 12 ul mixture of NMO‐IgG or human‐IgG supplemented
   with human complement (hC) (volume ratio 3:2). Mice in the STING
   inhibitor‐treated group received daily H151 (750 nmol; 200ul; i.p.)
   treatment.

AQP4‐Specific T Cells Mouse Model Induction

   The AQP4‐specific T cells mouse model was established following
   methodologies from previous studies.^[ [204]^20 , [205]^44 ^] C57BL/6
   Aqp4^‐/‐ mice were immunized with 200 µg of AQP4 peptides p135‐153 or
   p201‐220 (GeneScript Synthesis) in Freund's complete adjuvants
   containing Mycobacterium tuberculosis H37Ra. Meanwhile, mice received
   intraperitoneal injection of 200 ng pertussis toxin (Ptx) (List
   Biological) on days 0 and 2 after immunization. Draining lymph nodes
   were dissected on day 10‐12 after immunization. Th17 polarizing
   conditions were cultured with corresponding AQP4 peptides
   (AQP4[p135‐153] or AQP4[p201‐220]) (10 µg ml^−1), IL‐23 (20 ng ml^−1;
   R&D Systems), and IL‐6 (20 ng ml^−1; R&D Systems) in complete RPMI 1640
   medium (Gibco) for three days. Donor mice were intravenously injected
   with 1.5‐2 × 10^7 cells per mouse, and PT solution was i.p. injected on
   the day of induction and 48 h later. Clinical monitoring included daily
   assessments of body weight and symptoms post‐model induction. Paralysis
   severity was scored using a canonical clinical scale: 0) healthy; 0.5)
   partial tail weakness; 1) complete loss of tail tone; 2) partial
   hind‐limb paralysis; 2.5) Paralysis of one hind limb; 3) complete
   hindlimb paralysis; 4) forelimb paralysis; and 5) moribund/dead. Mice
   in the STING inhibitor‐treated group received daily treatments of H151
   (750 nmol; 200ul; i.p.).

Bulk RNA Sequencing Analysis

   Brain tissues were enzymatically digested following established
   protocols.^[ [206]^45 ^] Following model induction, mice were
   euthanized humanely, and brain tissues were collected and processed.
   Tissues were minced and incubated with collagenase IV and
   deoxyribonuclease at 37 °C for 30 min. After removing myelin debris
   through centrifugation using 30% Percoll, single cells were suspended
   in PBS supplemented with 2% FBS. Microglia isolation from the NMO‐IgG
   injection mouse model was achieved by utilizing flow cytometry sorting
   (FACS Aria III, BD Biosciences, San Jose, CA, USA). Libraries prepared
   from these samples underwent sequencing on an Illumina NovaSeq
   platform.^[ [207]^17 ^] Count data were generated using featureCounts,
   and the expression profiles of cGAS‐STING‐IFN‐I, disease‐associated
   microglia, and chemokine genes were normalized and visualized through
   the heat map.

Transmission Electron Microscopy (TEM)

   Transmission electron microscopy (TEM) samples were processed as
   described previously.^[ [208]^17 ^] Briefly, perfusion sampling was
   carried out 5 days after the induction of the NMO‐IgG model. First,
   20 ml of PBS was perfused, followed by 30 ml of fixative (2.5%
   glutaraldehyde, 4% paraformaldehyde in phosphate buffer; 0.1 M). Then,
   it was rinsed with phosphate buffer PBS and kept overnight in PBS,
   followed by fixation with 2% osmium tetroxide. The samples fixed with
   osmium tetroxide were dehydrated and embedded in epoxy resin.
   Ultra‐thin sections were stained with uranyl acetate and lead citrate,
   and the samples were examined using a Hitachi HT7800 transmission
   electron microscope at an accelerating voltage of 60 kV. Damaged
   mitochondria were defined as those having lost cristae and/or disrupted
   outer and/or inner mitochondrial membranes.^[ [209]^19 ^] The number of
   damaged mitochondria was manually counted using ImageJ software.

Immunofluorescence Staining

   Brain and spinal cord tissues were fixed in OCT compound (Tissue‐Tek,
   Miles) and sectioned at 8um (Leica CM1850, Leica Instruments) for
   immunofluorescence study. The sections were incubated with following
   primary antibodies: rabbit anti‐TMEM119 antibody (1: 500, Abcam,
   ab209064), mouse anti‐GFAP antibody (1: 1000, Abcam, ab4648), goat
   anti‐IBA1 antibody (1: 1000, Abcam, ab5076), rat anti‐C3 antibody (1:
   200, Abcam, ab11862), mouse anti‐C1q antibody (1: 200, Abcam, ab71940),
   mouse anti‐DNA antibody (1:200, Progen, AC‐30‐10), rabbit anti‐pIRF3
   antibody (1: 100, CST, 29 047), rabbit anti‐pTBK1 antibody (1: 100,
   CST, D52C2). TMEM119 and IBA1 were utilized for the identification of
   microglia, whereas GFAP was employed to label astrocytes. After
   overnight incubation at 4 °C, sections were washed and exposed to the
   appropriate corresponding secondary antibodies. Three to five
   corresponding sections were utilized for staining per animal, and
   fluorescence signal detection was conducted using the Leica TCS SP5
   laser scanning. The data were analyzed quantitatively using the ImageJ
   software. The cell count proportion was determined by the proportion of
   targeted cells within the lesion site at 20× magnification relative to
   the total number of cells in that area. Meanwhile, for assessing the
   DNA accumulation and phosphorylation level of the IRF3 or TBK1 in
   microglia, the mean fluorescence intensity (MFI) value of IBA1^+ cells
   within a region of interest (ROI) of the same size was calculated at
   40× magnification. Mean values from different sections of each animal
   were used for statistical analysis of differences.

Histology Staining

   The spinal cord sections were prepared as described for the
   immunofluorescence staining methodology. Hematoxylin & Eosin (H&E)
   staining of spinal cord sections was conducted to visualize
   inflammatory infiltration (Beijing Solarbio Science & Technology Co,
   G1121), and Luxol Fast blue (LFB) staining (Abcam, ab150675) was
   performed to visualize the demyelination of the spinal cord. Staining
   was conducted according to the manufacturer's protocols. Inflammation
   and demyelination scores were quantified from three different spinal
   cord sections as previously reported.^[ [210]^46 , [211]^47 ^]

Quantitative PCR Analysis

   Brain and spinal cord tissues were enzymatically digested as previously
   described, and CD11b^+ cells were isolated by incubating the
   single‐cell suspension with PE anti‐CD11b antibody (Biolegend, M1/70)
   and subsequently with anti‐PE microbeads (Miltenyi, 130‐048‐801). After
   sorting CD11b^+ cells, reverse transcription (RT) was performed using
   EZ‐press Cell to cDNA Kit PLUS (EZBioscience), according to the
   protocol provided by the manufacturer. Gene expression was detected by
   real‐time PCR with FastStart Universal SYBR Green Master (Roche) on CFX
   connect (Bio‐Rad). The primers are listed in Table [212]S3 (Supporting
   Information). Relative expression level was normalized by β‐actin as a
   housekeeping gene and calculated by the 2_ΔΔCt method. The data were
   presented as fold change relative to control samples.

Magnetic Resonance Imaging Scanning

   The lesion volumes of the NMO‐IgG intracerebral injection model were
   evaluated using a 9.4‐T small‐animal MRI scanner. T2‐weighted images
   were obtained for lesion detection with the following parameters: a
   repetition time (TR) of 2500 ms, an effective echo time (TE) of 33 ms,
   a resolution of 0.078 × 0.078 mm, two signal averages, and an imaging
   duration of 2 min and 40 s. Twenty‐four slices with a thickness
   measuring 0.5 mm were obtained using a field of view measuring 2.0 cm ×
   2.0 cm and a matrix size of 256 × 256. The MRI data analysis was
   performed utilizing ImageJ and Radiant DICOM Viewer software.

Flow Cytometry and Tetramer Analysis

   Single‐cell suspensions from mouse CNS, lymph node, or spleen tissues
   were prepared as described previously. Spleens were processed through a
   70‐µm cell strainer, and red blood cells were lysed using ACK lysis
   buffer. In the case of blood samples, red blood cells were lysed by
   adding ACK lysis buffer at room temperature for 15 min. Following
   centrifugation and PBS washing, cells were resuspended in a solution
   containing 1% bovine serum albumin (BSA) for subsequent antibody
   staining. In the intracellular cytokine staining experiment, a cell
   stimulation cocktail and protein transport inhibitor (eBioscience,
   00‐4975‐93) was employed to stimulate the cells, followed by the
   addition of corresponding antibodies for incubation under
   light‐shielded conditions. The following antibodies were utilized in
   the staining process: anti‐mouse CD45 antibody (Clone: 30‐F11),
   anti‐mouse/human CD11b antibody (Clone: M1/70), anti‐mouse Ly‐6G
   antibody (Clone: 1A8), anti‐mouse Ly‐6C antibody (Clone: HK1.4),
   anti‐mouse CD86 antibody (Clone: GL‐1), anti‐mouse CD3 antibody (Clone:
   17A2), anti‐mouse CD4 antibody (Clone: GK1.5), CD8a antibody (Clone:
   53‐6.7), anti‐mouse/human CD45R/B220 antibody (Clone: RA3‐6B2),
   anti‐mouse IFN‐γ antibody (Clone: XMG1.2,), anti‐mouse IL‐17A antibody
   (Clone: TC11‐18H10.1), Anti‐mouse IL‐6 antibody (Clone: MP5‐20F3), and
   7‐AAD Viability staining Solution (420 404). Antibodies were conjugated
   to fluorescent tags, including FITC, PE, PerCP‐Cy5.5, allophycocyanin
   (APC), PE/Cyanine7, APC/Cyanine7, and Brilliant Violet 421 (BV421). All
   antibodies were purchased from BioLegend. The AQP4 antigen‐specific T
   cells were detected by tetramer staining. The I‐A^b‐AQP4(205‐215)
   tetramer was used. Briefly, the single‐cell suspensions of the CNS
   tissues were stained with the I‐A^b‐AQP4(205‐215) tetramer at room
   temperature for 2 h, then washed with PBS, and anti‐mouse CD3 and CD4
   antibodies were added and incubated at 4 °C for 30 min. Flow cytometry
   assays were performed using BD FACS Aria III. The definition of
   different immune cell subsets in mice was described as follows: CD4^+ T
   cell: CD45^hi, CD11b^‐, CD3^+, CD4^+; CD8^+T cells: CD45^hi, CD11b^‐,
   CD3^+, CD8^+; B cells: CD45^hi, CD11b^‐, CD3^‐, B220^+; microglia:
   CD45^int, CD11b^+; myeloid cells: CD45^hi, CD11b^hi; neutrophils:
   CD45^hi, CD11b^hi, Ly6g^+. During the statistical analysis of various
   immune cell types in the CNS, experiments were conducted by mixing
   brain and spinal cord tissues from each mouse together to perform
   relative cell counting statistics for each specific cell type within a
   population of 10 000 live cells. The proportion of IL‐17A positive
   cells among CD4^+ T cells and the proportion of IL‐6 positive cells
   within B cells were quantified, and the gating strategy for cytokine
   experiments was established using fluorescence minus one (FMO) control.
   For the analysis of AQP4 antigen‐specific T cells, only T cells that
   are double positive for anti‐PE and anti‐APC fluorescence are
   identified as antigen‐specific T cells, and the proportion of these
   cells within all living cells was calculated. Flowjo 10.6.2 software
   was used for data analysis.

In Vitro IFN‐I Stimulation of AQP4‐Specific T Cells Assay

   Ten days after immunizing Aqp4^‐/‐ mice with AQP4[p201‐220], the
   draining lymph nodes were acquired and processed to obtain lymphocytes.
   The lymphocytes were then resuspended in complete RPMI‐1640 medium and
   seeded at a density of 3 × 10^5/200 µl in 96‐well plates pre‐coated
   with anti‐CD3 and anti‐CD28 antibodies (5 µg ml^−1 each, all from
   BioLegend, clones 145‐2C11, 37.51). The cells were cultured under three
   conditions: The culture groups comprised the control group: Baseline
   culture with CD3/CD28 co‐stimulation alone, Th17 polarization
   conditions: AQP4[p201‐220], 10 µg ml^−1; IL‐23, 20 ng ml^−1; and IL‐6,
   20 ng ml^−1, and Th17 polarization conditions with additional IFN‐I
   stimulation: IFNA1 (MCE, HY‐P700183AF), IFN‐β (MCE, HY‐[213]P73130),
   both at 100 U ml^−1. After 3 days of in vitro culture, the proportions
   of chemokine receptor CXCR3^+, CCR5^+, CCR6^+ cells and IL‐17A^+,
   IFN‐γ^+ cells in CD4^+ T cells were detected by flow cytometry. For the
   detection of antigen‐specific T cells, I‐A^b‐AQP4(205‐215) tetramer was
   employed for flow cytometry following 5 days of in vitro culture. Cells
   were treated with 0.7 U/ml neuraminidase (Sigma‐Aldrich, N3001) and
   10 nm dasatinib (Selleck) at 37 °C and 5% CO[2] for 30 min prior to
   staining, and subsequently stained with I‐A^b tetramers for 2 h at room
   temperature with repeated resuspension. Data analysis was performed
   using Flowjo 10.6.2.

Retroviral Transduction, and Adoptive Transfer

   AQP4[p201‐220]‐specific T cells were isolated from immunized Aqp4^‐/‐
   mice as previously described. Subsequently, the lymphocytes were
   activated with anti‐CD3 and anti‐CD28 antibodies (5 µg ml^−1 each) in
   complete RPMI 1640 medium for 48 h. Activated lymphocytes were
   transduced with retroviruses encoding either green fluorescent protein
   (GFP) with scramble shRNA (shRNA‐negative control), or GFP with mouse
   Sting‐targeting shRNA for 72 h. Transduced lymphocytes were then
   polarized under Th17 conditions (10 µg ml^−1 AQP4[p201‐220],
   20 ng ml^−1 IL‐6, and 20 ng ml^−1 IL‐23) for 3 days. For the cells
   utilized for adoptive transfer, after two washes, they were adoptively
   transferred into recipient WT mice through the vein at a dose of
   1 × 10^7 cells per mouse. PT was administered intraperitoneally at
   transfer and 48 h post‐transfer. Meanwhile, for the cells directly
   employed for detection after in vitro culture, they were processed
   using the flow cytometry method previously described.

Primary Microglia and CD4^+ AQP4‐Specific T Cells Co‐Culture Assay

   Primary microglia were isolated from WT or Sting^‐/‐ mice brain tissue.
   The CNS tissues were surgically extracted and incubated with
   collagenase IV at 37 °C for 30 min. Following myelin removal using 30%
   Percoll gradient, microglia (CD45^intCD11b^+) isolation was achieved
   through FACS. CD4^+ T cells were FCAS sorted from AQP4[p135‐153]
   immunized Aqp4^‐/‐ mice. Primary microglia (5 × 10^4/50 µl) were
   co‐cultured with AQP4‐specific T cells (1 × 10^5/200 µl) per well with
   or without corresponding AQP4[p135‐153] (20 µg ml^−1) stimulation at
   37 °C and 5% CO[2] for 3 days. Flow cytometry was employed for the
   quantitative assessment of IL‐17A and IFN‐γ positive cell proportions
   within CD4^+ T cells, as previously outlined. Data analysis was
   performed using Flowjo 10.6.2.

Carboxyfluorescein Succinimidyl Ester (CFSE) Assay

   CD4^+ T cells were sorted using FACS from AQP4[p135‐153] immunized
   Aqp4^‐/‐ mice, as previously described. Primary microglia
   (CD45^intCD11b^+) and monocytes (CD45^+CD11b^+Ly6c^+) from CNS or
   spleen were aseptically sorted by FACS from WT or Sting^‐/‐ mice. The
   CD4^+ T cells (10 × 10^6 cells ml^−1) suspensions were prepared and
   incubated with 0.5 µm CFSE away from light at room temperature for
   5 min. Then, CFSE labeling was stopped by adding a 5× excess volume of
   cold complete RPMI medium. Cells labeled with CFSE were seeded at 1 ×
   10^5/200 µl per well in 96‐well plates and the above primary microglia
   or monocytes were co‐cultured with the CFSE‐labelled CD4^+ T cells at 5
   × 10^4/50 µl (at a ratio of 1:2) per well with or without
   AQP4[p135‐153](20 µg ml^−1) stimulation at 37 °C and 5% CO2 for 3 days.
   CFSE‐labeled cells were collected and washed with PBS, and cells were
   resuspended in 1% BSA/PBS and stained with anti‐mouse CD3 and CD4
   antibodies and 7‐AAD at 4 °C in the dark for 30 min. Flow cytometry was
   performed, and data were analyzed using FlowJo 10.6.2 software.

Statistical Analysis

   The experimental data were presented as Mean ± s.e.m. Statistical
   comparisons between two groups were performed using an unpaired
   Student's t‐test, with the Mann‐Whitney test applied for non‐parametric
   data. ANOVA with Tukey's post‐test was used for multiple group
   comparisons, and the Kruskal‐Wallis test with Dunn's post‐test was used
   for non‐parametric data. Statistical significance was determined at the
   following p‐values: *P < 0.05; **P < 0.01; ***P < 0.001. All
   experimental data analysis and statistical evaluation were performed
   using GraphPad Prism 9 software.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   T.‐X.Z., X.Y., and X.G. contributed equally to this work and are
   co‐first authors. C.Z., F.‐D.S., A.L., and Z.K. formulated and designed
   the study. T.‐X.Z., X.Y., X.G., X.D., Y.L., and Z.H. performed the
   experiments. T.‐X.Z., X.G., X.D., X.L., N. S., Y.L., Z.H., and D.J.
   were involved in data collection, T.‐X.Z., X.Y., X.G., Z.L., and C.Z.
   analyzed all the results. T.‐X,Z., F.‐D.S., and C.Z. wrote the
   manuscript. All authors critically reviewed the manuscript.

Supporting information

   Supporting Information
   [214]ADVS-12-e00942-s001.pdf^ (5.8MB, pdf)

Acknowledgements